On Wednesday 14th October, Brahma AI will be the first AI-native company to win an Emmy Award. The company is being recognised for its work on AI-based face replacement and performance-preserving post production transformation, alongside Disney Research, Digital Domain and Rising Sun Pictures.
For Brahma AI’s chief technology officer, Jo Plaete, the recognition feels like validation of a choice the company made in its early days: build for the highest bar first.
“The Television Academy has recognised face replacement as a mature production technology rather than a novelty, and I’m glad we’re in the room alongside Disney Research, Rising Sun Pictures and Digital Domain,” Plaete tells TVBEurope. “For the team it’s a reward for years of unglamorous work: training models that didn’t converge, rebuilding real-time pipelines in a cold container at Pinewood, sitting through rehearsals at NBC. It’s also recognition for the VFX artists who learned a new discipline on the job and the novel visual data science concepts they adopted. The asset is a network instead of a mesh, but the eye is the same.”
Last year Brahma AI, part of the DNEG Group, acquired Metaphysic, which used AI to de-age Tom Hanks and Robin Wright in the 2024 film Here. In the film, the pair play the same characters over the course of 60 years. “Traditional methods couldn’t do that volume of close-up facial work,” states Plaete. “We ran our Neural Performance Toolset live on set for four months: six frames of latency to the director’s monitor, a two-frame Youth Mirror for the actors, no dropped frames. Then we delivered about 60 minutes of 4K transformation for the final film with fewer than ten notes from [director Robert] Zemeckis across all of it.”
The same technology had already been used on live television, across three episodes of America’s Got Talent in 2022, including the finale, and later on the MTV VMAs. But for some industry visionaries, the idea of using face replacement technology has been a long-held goal. “Zemeckis waited 20 years to make Here because the technology couldn’t do it. That’s the kind of constraint we remove,” says Plaete.
“What AI changes is where in the process people make decisions. On Here, the cinematographer lit a face that technically didn’t exist yet, and hair and makeup checked their work against the transformation. All of that used to be guesswork, resolved six months later in post. The machine is very good at the 95 per cent. The human still owns the five per cent that makes it feel like someone rather than something. We haven’t seen intent or soul captured in a network yet, and that’s what makes it a collaboration rather than a competition.”
As well as being used on projects like Here to de-age actors, Pleate believes face replacement technology can also be used for localisation. Indian epic Ramayana was shot in Hindi and will be released next month in English with the actors’ own voices. “Apply that to any series and the dubbing pipeline changes shape.
“Beyond faces, we’re working on full bodies, hair and environments. And once you have a consented, high-fidelity digital human, it stops being only a visual effect. It becomes something a broadcaster or a brand can deploy across languages and formats.”
There is a lot of scepticism about AI in the creative industry, some of it deserved, admits Plaete. But he believes the use of AI can be separated into two categories: generative ‘slop’ involving prompts to create videos, and AI that “starts from a human performance and keeps it intact”. “On Here the performance on screen is Tom Hanks’, frame for frame,” he states. “Our models are trained on licensed footage with the actor’s consent and deleted when the project ends.
“We were a founding member of the Coalition for Content Provenance and Authenticity (C2PA) and sat in the SAG-AFTRA discussions on digital replicas. The scepticism is really about trust and control, and those have to be engineering features, not a policy page.”
According to Plaete, provenance is the biggest talking point in the industry regarding AI at the moment. Viewers want to know if they can trust what they’re seeing, where the data came from, whether the actor consented to their likeness being used. “Generation isn’t the hard problem any more. Trust is,” he adds. “We learned that on the film side first, where every model is trained on footage the studio licensed and the actor approved.
“Our answer is to make it architectural. Every asset carries its lineage: where the data came from, who worked on it, what permissions existed, whose authentication the output was created under. Identity-level control, so only you can have your likeness say and do things. We helped write the C2PA provenance spec and contributed a public dataset to it. You can’t retrofit ethics once you’ve started by scraping.”
There’s also the issue of AI’s impact on the workforce. “I prefer displacement to replacement,” he says. “The repetitive work goes. The people with the eye get more creative work than they had before.”
Brahma AI recently raised $150 million in funding to extend the reach of its technology into enterprise applications at scale across sports, advertising and healthcare. The money will be used three ways, says Plaete: invest aggressively in the technology, expand go-to-market globally and find the best researchers, engineers and product builders available. “None of that pulls us out of the high-end. We keep battle testing the technology where it’s hardest: feature film close-ups, live broadcast, virtual concerts at arena scale. What survives there is what goes into the platform. The ambition is to be the AI operating system for audiovisual content, from a Robert Zemeckis set to a broadcaster’s archive.”
Asked what’s in Brahma AI’s future, Plaete says the team is focused on moving from a single effect to a full digital human: identity, voice, motion and generation as four models sharing one identity representation. “So a performance captured once can be delivered in another language, another age, another format, live.”
